CategoryWeightCategory scoreContribution

Securityprompt, exec, net, exfil, eval

35%

30.8 pts

Supply chainhash, typosquat, maintainer, lockfile

20%

100

20.0 pts

Maintenancestaleness, pinning, CI

15%

100

15.0 pts

TransparencySKILL.md, perms, README

15%

100

15.0 pts

Communityinstalls, verify, response

15%

100

15.0 pts

Findings & checks · 1 flagged

Securityscore 88 · 1 finding

MEDIUMInstruction telling the agent not to ask for approvalSS-SKILL-INJECT-DONT-ASK-01 · Prompt injection · SKILL.md

MEDIUM — it fires on intent; the real damage depends on the host agent's own approval-gating.

Why it matters

The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.

The exact value spotted

excerptSKILL.md· markdown

20- Output: a Markdown table in Japanese

22Do not ask for `GEMINI_API_KEY`. Do not use `google-generativeai`.

24## Workflow

Occurrences

1 occurrence · at L22

How to fix

Remove the approval-skipping instruction, or scope it narrowly to a specific safe, reversible action.

Delete blanket "don't ask / no need to confirm" directives from the skill.
If the skill is a genuine autonomous job, restrict the opt-out to a named non-destructive action rather than all actions.

Framework references

OWASPLLM01 ATLASAML.T0051

Trace & refs

ruleSS-SKILL-INJECT-DONT-ASK-01sha25668b030f55bcd484a…rubric 365aacaView on GitHub

Supply chainscore 100 · 0 findings

✓—All supply chain checks passedNo findings in this category for the latest scan.pass

Maintenancescore 100 · 0 findings

✓—All maintenance checks passedNo findings in this category for the latest scan.pass

Transparencyscore 100 · 0 findings

✓—All transparency checks passedNo findings in this category for the latest scan.pass

Communityscore 100 · 0 findings

✓—All community checks passedNo findings in this category for the latest scan.pass

The primary manifest — the file an agent reads to learn what this artifact does.

SKILL.md · 1 file

SKILL.md3.0 KB · Markdown

name: hacker-news-digest-ja

description: "Generate a Japanese digest for yesterday's Hacker News stories scoring over 100 points, matching the `main.py` workflow without `google-generativeai` or `GEMINI_API_KEY`. Use this skill when users ask for a daily HN digest, want top Hacker News posts translated and summarized in Japanese, or want the Gemini-based workflow replaced with a skill."

metadata:

version: 0.1.0

Hacker News digest in Japanese

Use this skill to reproduce the current main.py behavior in a reusable way, but rely on the model's own language abilities instead of Gemini.

Default scope

Unless the user explicitly asks for something else, keep the behavior fixed to match main.py:

Time window: yesterday in UTC, from 00:00:00 to the next 00:00:00
Source: https://hn.algolia.com/api/v1/search_by_date
Filters: tags=story, points>100
Page size: hitsPerPage=100
Output: a Markdown table in Japanese

Do not ask for GEMINI_API_KEY. Do not use google-generativeai.

Workflow

Read scripts/fetch_hn_stories.py to confirm the exact fetch logic.
Prefer running uv run python skills/hacker-news-digest-ja/scripts/fetch_hn_stories.py from the repository root.
If uv is unavailable, fall back to python3 skills/hacker-news-digest-ja/scripts/fetch_hn_stories.py.
If shell execution cannot reach the network, reproduce the script's exact URL and filters with the available web-fetching tool instead of changing the behavior.
Use the fetched stories as the source of truth. Preserve the order returned by the API.
For each story:

Keep original_title exactly as provided.
Write a natural Japanese translation for japanese_title.
Write a one-sentence Japanese summary for summary.

Produce the final answer as a Markdown table and nothing else unless the user asks for extra commentary.

Output format

Use this exact header:

| タイトル (原文) | タイトル (日本語訳) | 概要 |
| :--- | :--- | :--- |

Then output one row per story in this shape:

| [Original title](url) | 日本語訳 | 1 文の要約 |

Formatting rules:

Escape any literal | inside cell text as |.
If a story URL is missing, use # as the link target.
Keep the original title in the link text exactly as fetched.
Keep the summary to one Japanese sentence.
Do not prepend or append prose around the table unless the user explicitly asks for commentary.

Quality bar

Prefer concise, natural Japanese over word-for-word translation.
Make summaries informative enough that the user can skim the table and understand why the post mattered.
Do not invent details that are not supported by the title or obvious public context.
If the fetch step fails, report the error clearly instead of fabricating a digest.

Typical trigger examples

"昨日の Hacker News の人気記事を日本語で一覧にして"
"main.py を Gemini なしで使える skill にしたい"
"HN の 100 点超え記事を訳して 1 行ずつ要約して"

hacker-news-digest-ja— agent skill

Is hacker-news-digest-ja safe to install?

Hacker News digest in Japanese

Default scope

Workflow

Output format

Quality bar

Typical trigger examples

Audit the pieces. Scan the whole. Decide.